As AI-based weather prediction model emerges, its prediction skill for extreme weather draws great attention. This study assesses the skill of GraphCast in heavy precipitation prediction against rain gauge observations focusing on 10 heavy precipitation cases in South Korea. For comparisons, the Weather Research and Forecasting (WRF) model simulations with different horizontal resolutions of 27 (WRF27km), 9 (WRF9km), and 3 km (WRF3km) are used. When the skill in predicting general patterns of precipitation is evaluated using the normalized mean bias, normalized centered root-mean-square error, and spatial correlation coefficient, GraphCast shows better or comparable skill compared with WRF27km, WRF9km, and WRF3km. When the skill for different precipitation thresholds is separately examined using the equitable threat score (ETS), neighborhood ETS, and other contingency table-based metrics, GraphCast shows relatively large overprediction of light precipitation (≤ 20 mm) and underprediction of heavy precipitation (≥ 70 mm) compared with WRF27km as well as WRF9km and WRF3km. GraphCast reproduces observed topographic enhancements of precipitation over the Sobaek Mountains by 42

Below-cloud scavenging is an important removal process of particulate matter (PM) and its effects change with characteristics of rainfall as well as PM. This study examines below-cloud scavenging of fine and coarse PMs in Seoul, South Korea and its change with rainfall types using PM and Parsivel(2) disdrometer observation data during similar to 7 years. PM concentrations 1 h before and after rainfall events are compared for PM10, PM2.5, and PM2.5-10. The scavenging effects on the PM concentrations increase with increasing rainfall amount, and PM2.5-10 is more effectively scavenged than PM2.5. Both the increases in rainfall duration and rainfall intensity lead to the increase in scavenging effect but the rainfall duration plays a more important role in increasing scavenging effect than the rainfall intensity. To examine the change in scavenging effect with microphysical characteristics of rainfall, stratiform-type (type S) and convective-type (type C) rainfall events are grouped and compared. For similar rainfall amounts, the type S rainfall events exhibit longer duration, weaker intensity, and higher scavenging efficiency than the type C rainfall events. The higher scavenging efficiency of the type S rainfall events is associated with their smaller raindrop sizes, which lead to larger total volume swept by raindrops and higher collision efficiency between PMs and raindrops. Empirical relations between rainfall rate and scavenging coefficient are obtained for PM10, PM2.5, and PM2.5-10 concentrations in Seoul. The relations tend to yield smaller scavenging coefficients compared with the relations obtained from Beijing, China and Sado island, Japan.

The sensitivity of aerosol effects on orographic precipitation from deep convective clouds to mountain upslope steepness is examined using the Weather Research and Forecasting (WRF) model coupled with a bin microphysics scheme. During the early stage, the sensitivity resembles that of warm, shallow orographic convection as discussed in Part I. As time progresses, interactions between vigorously developed lower-layer clouds and upstream-extending upper-layer clouds become crucial for enhancing surface precipitation via melting and direct sedimentation of ice-phase particles such as graupel and hail. In the simulations with a symmetric mountain shape, higher aerosol number concentration enhances surface precipitation through stronger condensational latent heating and more active mixed-phase processes (freezing, Wegener-Bergeron-Findeisen process, and riming). Under asymmetric mountain shapes, however, the sensitivities are non-monotonic. In the steep upslope cases, fast liquid drop growth in the clean case and strong latent heating in the polluted case both support cloud development and enhance precipitation. In contrast, the control case produces less precipitation because its drop growth is slower than in the clean case and its latent heating is weaker than in the polluted case. As a result, cloud interaction is suppressed. In the gentle upslope cases, the control case shows the most precipitation due to sufficient droplet supply and latent heating which promote vertical growth and cloud interaction. The clean case lacks enough droplets, while the polluted case suffers from weak convection despite strong aerosol-induced heating. Consequently, both cases exhibit suppressed cloud interaction and mixed-phase processes.

Various scales of turbulent coherent structures (TCSs) individually and collectively contribute to pollutant dispersion in urban areas. Although it is well known that surface heating significantly modifies urban turbulent flows, its impact on TCSs with different scales and their interactions still needs further investigations. In this study, we examine the impact of surface heating on different scales of TCSs and their associated turbulent pollutant flux over an idealized building array. Two large-eddy simulations with and without surface heating are conducted using the PArallelized Large-eddy simulation Model (PALM). The spectral proper orthogonal decomposition (SPOD) and the amplitude modulation (AM) analysis are applied to extract different scales of TCSs and to examine their interactions, respectively. The surface heating enhances low-speed streaks at and above the roof level while it leads high-speed regions to be less organized into streaks. The surface heating preferentially enhances large-scale coherent structures and therefore substantially increases the relative importance of large-scale coherent structures for roof-level TKE and turbulent pollutant flux compared with mid-scale and small-scale coherent structures. The surface heating also enhances the modulation of small-scale turbulent momentum and pollutant fluxes by large-scale coherent structures above the roof level. This enhancement of the modulation mainly originates from the strengthened large-scale upward motions. This study emphasizes the necessity of applying modal decomposition and scale-interaction analysis together to better understand both scale-dependent characteristics of TCSs and interactions among TCSs with different scales.

This study quantifies and compares the relative importance of tree-cast shadows and transpiration in mitigating urban extreme heat through online numerical simulations using the Weather Research and Forecasting (WRF) model coupled with an urban canopy model that represents street trees and a factor-separation method. During the extreme heat event, tree-cast shadows and transpiration by street trees with a crown radius of 1 m reduce urban canyon air temperature by 0.43 ± 0.06 °C and 1.15 ± 0.09 °C in the daytime, respectively, the tree transpiration effect being much greater than the tree-cast shadow effect. The warming associated with the interception of direct shortwave radiation by street trees is greater than the cooling associated with tree-cast shadows in the daytime. In the nighttime, the mean cooling effects due to tree-cast shadows and transpiration are 0.09 ± 0.02 °C and 0.15 ± 0.03 °C, respectively, the two cooling effects being comparably small. The relative importance of the two effects is found to be similar under differing urban geometry and tree-crown size. This study provides quantitative evidence highlighting the importance of managing tree soil moisture to sustain tree-induced cooling benefits.

Turbulent coherent structures (TCSs) play important roles in momentum, pollutant, and heat transports. While TCSs over urban areas have been widely investigated assuming neutral conditions or constant surface heat fluxes, TCSs over urban surfaces heated/cooled by realistic radiative processes are little investigated yet. This study examines TCSs and their roles in pollutant dispersion and heat transport over radiatively heated/cooled urban surfaces using the PArallelized Large-eddy simulation Model (PALM) including radiative physics. A cubical building array is considered. The radiative processes result in heterogeneous temperature distributions on roof, road, and wall surfaces during daytime and nighttime. During daytime, low-speed streak appears up to well above the roof level while high-speed streak is found only near the roof level. During nighttime, the sizes of TCSs are markedly small and streaks are not found. The spectral proper orthogonal decomposition (SPOD) analysis is conducted to identify coherent structures of different scales and quantify their contributions to pollutant dispersion and heat transport. During daytime, coherent structures of various scales appear near and above the roof level. Relatively large (small) coherent structures exhibit relatively large (small) contributions to turbulent kinetic energy (TKE) but show relatively small (large) contributions to vertical turbulent pollutant and heat fluxes at the roof level. During nighttime, coherent structures are comprised of small irregular vortices. Coherent structures having relatively large (small) contributions to TKE show relatively large (small) contributions to vertical turbulent pollutant and heat fluxes above the roof level but exhibit relatively small (large) contributions at the roof level.

This study examines the synergies between urban heat islands (UHIs) and heat waves (HWs) and their quantitative effect on urban heat stress in Dhaka, Bangladesh under HWs. For this, total 51 days in the pre-monsoon seasons of 2020, 2021, and 2022 are simulated using the Weather Research and Forecasting model. During the HW days, the synergies between UHIs and HWs are weak in the daytime (0.07 degrees C) and pronounced in the nighttime (1.28 degrees C). Due to HWs, the heat storage and the release of stored heat increase greatly in the urban area, which is responsible for the pronounced synergies in the nighttime. To quantitatively evaluate the effect of the synergies on urban heat stress, individual effects of HWs, UHIs, and their synergies on urban heat stress are quantified and the normalized relative impact index is defined. The HW effect plays the most important role in daytime urban heat stress under HWs, and the UHI effect is most important for aggravating nighttime urban heat stress under HWs. The synergistic effect on urban heat stress is minor in the daytime, whereas it plays roles in worsening nighttime urban heat stress, its influence being 37-44% of the UHI effect.

Abstract Convectively forced gravity waves (CGWs) play important roles in the atmosphere. CGW drag parameterizations for use in large-scale models typically represent subgrid-scale convection as a single convective source. This study uses an analytic framework to investigate how horizontal organization of multiple convective cells modifies the vertical flux of horizontal momentum. It is shown that the per-cell momentum flux induced by multiple convective cells is expressed as the momentum flux induced by a single convective cell times [1 + J n (β, δ)]. Here, J n (β, δ) is the nonlinear modification factor arising from mutual interactions among distinct convective cells, where β is the ratio of compensating-cooling width to heating-core width, δ is the nondimensional spacing between the horizontal centers of convective cells, and n is the number of cells. Closely packed deep convective cells with broad cooling (large β and small δ) act to enhance the per-cell momentum flux relative to a single convective cell, whereas more widely spaced convective cells act to reduce it. As the number of convective cells increases, the region of positive J n in the parameter space of β and δ shrinks toward smaller δ. These results indicate that purely geometric properties of convective organization (β, δ, and n) systematically modify CGW momentum flux. The proposed nonlinear modification factor can be easily incorporated into existing CGW drag parameterizations.

Weather and climate changes due to urbanization are of great concern. This study examines urban effects on thermal and wind environments in the Dhaka metropolitan area, Bangladesh. For this, simulations for a case of hot days with weak synoptic forcing are performed using the Weather Research and Forecasting (WRF) model. The differences between the urban and no-urban simulations are analyzed. In the urban simulation, the daytime sensible (latent) heat flux is considerably increased (reduced) and convective activities are enhanced. The nighttime (0000-0500 LST) urban heat island (UHI) is much stronger than the daytime (1200-1700 LST) UHI. In the daytime, the UHI effect on local winds is more important than the urban surface roughness effect. In the nighttime, the relative importance of the UHI and urban surface roughness effects differs depending on the region for given prevailing winds. Impacts of increases in anthropogenic heat and urban size are examined. As the anthropogenic heat increases, the UHI and the UHI effect on local winds strengthen. As the urban size increases, the UHI and the UHI effect on local winds strengthen and the surface roughness effect appears in wider areas. This study provides further insights into urban effects on local winds.

This study evaluates the performance of a recently developed Lagrangian stochastic model (LSM), which solves the transport equations for turbulence probability density functions (PDFs), for simulating the unstable atmospheric surface layer (ASL). The simulated statistics are compared with the Monin-Obukhov similarity theory (MOST) predictions for mean gradients, standard deviations, turbulent Prandtl number, turbulence kinetic energy budgets, and turbulence PDFs. The LSM successfully captures many aspects of ASL structure and turbulence characteristics, particularly the mean gradients and standard deviations of potential temperature, specific humidity, and vertical velocity, which align closely with MOST predictions. However, the model shows limitations in reproducing the stability dependency of mean gradients of horizontal wind speed. Additionally, the model underestimates the probability of high vertical velocity fluctuations, leading to underprediction of turbulent transport and thermal convection initiation. The simulated turbulent Prandtl number shows dependency on stability but is generally weaker than MOST predictions. The study identifies that the parameterizations of dissipation, pressure redistribution, and pressure transport need to be enhanced to ensure that they have correct stability dependency, in order to improve the model’s accuracy in simulating the ASL.

Precipitation systems passing over cities can be modified by urban impacts. Compared with urban impacts on less-organized convective precipitation systems, urban modification of highly-organized or strongly-forced precipitation systems remains less understood. In this study, we examine urban impacts on a cold-frontal precipitation system passing over the Seoul metropolitan area, South Korea. For this, three sets of urban ensemble simulations with diurnal-mean anthropogenic heat intensities and urban momentum roughness lengths of (0 W m-2, 1 m), (45 W m-2, 2 m), and (90 W m-2, 3 m) and a set of nonurban ensemble simulations are performed. Comparisons with observations show that the characteristics and passage of the cold front are acceptably reproduced in all sets of the ensemble simulations. Precipitation is enhanced by about 20-40 % downstorm of Seoul in the urban ensemble simulations, with marginal delays in the movement of the precipitation system. The precipitation enhancement is hard to be discerned by a comparison between a single pair of urban and nonurban simulations, since the urban-induced precipitation enhancement is obscured by strong precipitation anomalies arising from different locations of precipitation cores in individual simulations. The downstorm precipitation enhancement is attributed to precipitation intensity increases in narrow cold-frontal rainband, which results from intensified frontal updrafts due to the increases in cross-frontal vorticity. The warm urban-heated air trailing the cold front after the frontal passage results in the increases in cross-frontal vorticity by decreasing temperature gradient at the frontal zone.

Urban canopy models (UCMs) help to better understand urban climatic phenomena including surface urban heat islands (SUHIs). Previous studies employing UCMs investigated the causes of SUHIs by analyzing surface energy fluxes balanced for an urban conceptual volume (bulk approach). In this approach, since the urban surface energy fluxes are represented at the top surface of the conceptual volume, the representative urban surface temperature is the radiative surface temperature inverted from emitted longwave radiation there. Meanwhile, the thermodynamic urban surface temperature is represented by the surface temperatures of urban facets (roofs, roads, and walls), the causes of SUHIs based on them being yet to be thoroughly investigated. Here, we examine the causes of SUHIs using facet surface temperatures simulated by a UCM. For this, the simulated surface energy fluxes at individual facets are area-weighted averaged and analyzed (facet approach). Two-dimensional idealized simulations are conducted, roughly representing mid-latitude hot and dry summer conditions. In both approaches, the primary cause of daytime SUHI is less evapotranspiration in the urban area. The amount of net shortwave radiation averaged over facets (inside the urban conceptual volume) is smaller (larger) than that at the rural surface, being interpreted to weaken (intensify) the daytime SUHI in the facet (bulk) approach. The nighttime SUHI is attributed to larger urban heat storage in the bulk approach but the trapping of longwave radiation in the facet approach. This study suggests that the facet approach is physically more consistent than the bulk approach in investigating the causes of SUHIs.

AbstractThis study analyzed geodetic distribution about temporal characteristics in rainstorm (> 1 hour) observed at approximately 600 rainfall stations across Republic of Korea. Utilizing minute-scale precipitation data observed by rainfall stations from 2000 to 2022, independent rainstorm events separated from rainfall data per unit time (i.e., 10, 20, 30, and 60 minutes) and Inter-Event Time Definition (IETD) (i.e., 2, 3, 4, and 6 hours). The significant variations in rainfall characteristics are defined as the number of independent rainstorm events, rainfall duration (hour), amount (mm), and intensity (mm/hour) for quantifying the temporal characteristics across rainfall stations. We quantified temporal characteristics among rainfall characteristics observed by rainfall stations based on latitude and longitude. The number of independent rainstorm events varies significantly depending on unit time and IETD, and the occurrence of events was frequently observed in areas characterized by island features. The rainfall amount for independent rainstorm events obscured significant characteristics, excluding Halla Mountain on Jeju Island. The geodetic distribution for the duration and intensity per rainstorm event varied depending on the characteristics of the region (i.e., island, mountain, etc.). Based on these results, it was confirmed that certain temporal characteristics vary according to regional features. In future research, we intend to utilize this information to cluster rainfall stations based on temporal characteristics.Keywords: Independent Rainstorm Events, Temporal Characteristics, Geodetic Distribution, Regional Features, Republic of KoreaAcknowledgmentThis research was supported by Korea Environment Industry & Technology Institute (KEITI) funded by Korea Ministry of Environment (RS-2022-KE002032 and 2022003640001) and was also supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2022R1A4A3032838 and No. RS-2023-00250239).

The multi-physics ensemble method is a widely used method to represent the prediction uncertainty arising from model errors and has great potential to improve precipitation forecasts. This study performs multi-physics ensemble simulations of ten heavy precipitation cases in South Korea and evaluates the performance of the ensemble mean. The multi-physics ensemble is generated from 27 different combinations of cloud microphysics, planetary boundary layer, and radiation schemes. In the prediction of 24-h accumulated precipitation amount, the ensemble is underdispersive for most cases, indicating that the prediction uncertainty is only partly represented by the ensemble. The overall performance of the ensemble mean is better than that of any individual ensemble member. No individual ensemble member consistently shows good performance for every case. The root-mean-square error (RMSE) of the ensemble-mean prediction is smaller than the average RMSE of individual ensemble members for every case. The relative difference in RMSE exhibits a strong positive correlation with the spread-error ratio. In comparison with subset ensembles, the total ensemble shows the most stable performance. Among the three types of physics parameterization in the multi-physics suite, the cloud microphysics parameterization contributes the most and the radiation parameterization contributes the least to the ensemble spread and the ensemble-mean performance. Each physics parameterization scheme has tendencies to predict cloud and precipitation properties to be larger or smaller than those predicted by other schemes, which stem from differences in the choices of individual process parameterizations and the physics parameter values used. This systematic difference contributes to the ensemble spread of the multi-physics ensemble, which is a key factor for the ensemble-mean performance.

This study investigates the impacts of cool roofs on the urban heat island (UHI) and air quality in Dhaka, Bangladesh during an extreme heat wave event occurring in April 2021. A simulation with conventional roofs having an albedo of 0.2 and a simulation with cool roofs having an albedo of 0.8 are conducted using the Weather Research and Forecasting (WRF) model. Cool roofs reduce the 2-m temperature by 0.57 degrees C in the afternoon (1200-1700 LST) and cause the urban cool island in the daytime. In the afternoon, cool roofs reduce the planetary boundary layer height by 265 m and greatly suppress the urban breeze, reducing the 10-m wind speed by 0.8 m s- 1. As a result, the near-surface passive tracer (carbon monoxide) concentration increases by 45 ppb (52 %) in the afternoon. Cool-roof effects on the UHI and air quality are overall more pronounced in hotter areas. Cool roofs lead to statistically significant decreases in Humidex (-0.19), discomfort index (0.22), and heat index (-0.36 degrees C) in the afternoon, but all indices remain within the same stress levels. This suggests that additional measures such as urban greenery and other climate-sensitive urban designs are required along with cool roofs for an effective mitigation of urban extreme heat in Dhaka.

Recently, the frequency and severity of droughts have gradually increased due to extreme weather events and global warming. As the demand for drought management increases, field surveys and water supply are actively conducted in many countries. Given that such drought assessment and support require the consumption of labor and financial resources, the prioritization of essential agricultural areas has become a major topic for efficient decision-making in drought relief. In this study, we proposed a Principal Component Analysis (PCA) for selecting rural specialization districts across the 162 administrative regions of South Korea. Additionally, we aimed to investigate real cases of agricultural drought occurred in these regions by utilizing the survey of water supply measures derived from Ministry of Agriculture, Food and Rural Affairs. The research data comprised seven agricultural specialization factors, exemplified by agricultural workforce and infrastructure. First, we implemented singular decomposition method included in PCA process to represent the comprehensive trends of the agricultural specialization factors with maximum reflection. High value of principal component scores (PCS) estimated from PCA was interpreted as regions with high agricultural relevance. Lastly, the PCS were classified into different levels, defining top-ranking regions as rural specialization districts. Based on agricultural drought case studies from 2018 to 2021, it is expected that finding relative damage-prone areas and establishing appropriate drought responses will be feasible. Keywords: Principal Component Analysis, Rural Specialization Districts, Agricultural Specialization Factors, Principal Components Score Acknowledgement This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. RS-2023-00250239) and this research was supported by Korea Environment Industry & Technology Institute (KEITI) through Water Management Innovation Program for Drought (RS-2022-KE002032) funded by Korea Ministry of Environment. This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No.NRF-2022R1A4A3032838).

Turbulent coherent structures over urban surfaces are known to play important roles in turbulent momentum and heat transfers. However, their associations with pollutant dispersion are less known, especially when urban surfaces are heated or cooled. This study examines the effects of roof surface heat flux on turbulent coherent structures and pollutant dispersion in and above a cubical building array using the PArallelized Large-eddy simulation Model (PALM). Three simulations with roof surface heat fluxes of 0 (neutral), 0.1 (heating), and-0.1 K m s(-1)(cooling) are conducted. The turbulent coherent structures in the heating and neutral simulations well develop in the horizontal and vertical directions, exhibiting large correlations between the pedestrian level and roof level and between the roof level and above. On the other hand, in the cooling simulation, the sizes of turbulent coherent structures are small compared to those in the heating and neutral simulations and turbulent coherent structures exhibit no correlations between the roof level and above. To examine the effects of roof surface heat flux on turbulent momentum, pollutant, and heat fluxes, quadrant analysis is performed at the roof level. In all three simulations, the ejection and sweep events most frequently occur and they are accompanied by pollutant ejection and pollutant sweep events, respectively. The frequency of warm updraft (cold updraft) events is lowest in the heating (cooling) simulation, being associated with marginal changes in potential temperature with height below the roof level. To further examine the effects of roof surface heat flux on turbulent flow and pollutant dispersion during strong ejection and sweep events, flushing events and cavity eddy events are identified. During both flushing and cavity eddy events, vertical turbulent pollutant exchanges are accompanied by stronger (weaker) vertical turbulent flows in the heating (cooling) simulation than in the neutral simulation.

With growing urban population and expanding urban areas, the importance of understanding urban effects on precipitation keeps increasing. This study attempts to detect urban effects on precipitation in the Seoul Metropolitan Area (SMA), South Korea by analyzing hourly rain gauge data during 2005–2020. Precipitation events are categorized according to 850-hPa wind directions, and the precipitation increases from the upwind to downwind regions are examined for different duration and intensity classes of precipitation events. The downwind precipitation increase is largest in summer (39%), especially in August (64%). The August precipitation is analyzed in detail. Precipitation statistically significantly increases in Seoul for weak winds and 25–50 km downwind of the center of Seoul for westerly winds, and the precipitation increases are largest in the afternoon. For the precipitation increases, the increases in frequency and intensity of precipitation events are responsible. Short-duration and heavy precipitation events associated with small-sized precipitation systems initiated within the SMA are mainly responsible for the precipitation increases. The downwind precipitation increase also occurs for southwesterly, southerly, and southeasterly winds, but the increases are associated with large-sized precipitation systems.

In this study, the microphysical characteristics of snowfall in Seoul, South Korea and their changes with meteorological conditions are examined using about 6-year observation data from a Parsivel disdrometer. The snow particle size distribution (PSD) exhibits convex-down shapes, being better represented by gamma distributions than exponential distributions. As snowfall rate increases, the snow PSD broadens and its peak rises. The changes in gamma PSD parameters with snowfall rate differ between the mean PSD and 1-min PSDs. The volume-weighted mean diameter Dm much more rapidly increases with snowfall rate in comparison with Dm in Beijing, China and Pyeongchang, South Korea, suggesting the relative importance of aggregation in Seoul. 77

A single-column turbulence model for stratified atmospheric boundary layer (ABL), which solves the transport equations of turbulence probability density function (PDF) using a Lagrangian stochastic modeling (LSM) approach, is proposed in this study. This study adopts previously developed stochastic differential equations (SDEs) for particle velocity and temperature and extends the LSM to simulate inhomogeneous turbulence. The proposed LSM is tested for its ability to fully simulate statistics of inhomogeneous stratified turbulence. In the model, particles evolve by SDEs, and turbulence statistics are calculated by averaging the properties of particles. The model provides a full representation of turbulence PDF and simulates turbulent transport without any modeling assumption. The model performance is evaluated against large-eddy simulation (LES) results in the simulations of convective and stable ABL cases. For the convective ABL, LSM realistically simulates the entrainment process with the temperature and heat flux profiles that closely match with LES. The joint PDF simulated by LSM reproduces a curved and highly skewed shape, and some distinct features, like the asymmetric distribution of vertical velocity and the separation of the PDF in the entrainment zone, are simulated. LSM also reproduces the entrainment enhancement by wind shear in the simulation of sheared convective ABL. The LSM simulation of stable ABL predicts realistic turbulence intensity and mean field profiles, where Gaussian-like PDFs are simulated both in LSM and LES.